1. Technical Field
The present invention relates to eliminating false audio associated with Voice over Internet Protocol (VoIP) communications.
2. Brief Discussion of Related Art
In VoIP networks, an access VoIP gateway can perform a signal classification function to classify a signal as a voice signal or a dual-tone multi-frequency (DTMF) signal. DTMF signals use two frequencies to represent digits and symbols of a phone keypad. When any key on the keypad is pressed the two tones are generated based on a frequency pair associated with the key that is pressed. These tones can be decoded by a network component to determine which key was pressed. DTMF signals are typically not transmitted over a VoIP network as audio, but rather as a packet including a characterization of parameters associated with the DTMF signal which is transmitted over the VoIP network.
When the signal is classified as a voice signal, the voice signal is processed via a voice codec implemented by the access VoIP gateway in the VoIP network, and the output (e.g., an encoded voice signal) of the voice codec can be encapsulated using a Real-time Transport Protocol (RTP) audio packet that can be transmitted through the VoIP network to an egress VoIP gateway. The egress gateway decodes the encoded voice signal in the RTP audio packets into a form used by an interfacing non-VoIP network. When the signal is classified as a DTMF signal, the DTMF signal is not processed by the voice codec of the access gateway. Rather, the signal is characterized based on parameters of the DTMF signal. This characterization information can be encapsulated in an RTP DTMF packet that is transmitted through the VoIP network to the egress VoIP gateway. The egress VoIP gateway can generate a proper DTMF signal based on the characterization information included in the DTMF packet.
Conventional signal classification processes implemented by access VoIP gateways can misclassify signals. For example, the VoIP gateways can misclassify a portion of a DTMF signal as a voice signal. Such misclassification can be a result of, for example, biasing the classification process towards voice classification and/or a delay in switching from a voice processing mode to a DTMF processing mode. When the initial energy of the DTMF signal is misclassified, the initial energy of the DTMF signal can be processed via the voice codec before switching to DTMF transmit mode. This imperfection is typically referred to as DTMF “leakage” since the initial DTMF energy is misclassified as a voice signal. That is, instead of properly classifying the signal as a DTMF signal, some of the DTMF signal “leaks” into the voice processing process implemented by the codec in the access VoIP gateway.
The consequence of DTMF leakage is that the egress VoIP gateway first receives an audio packet carrying false audio and then receives the DTMF characterization information associated with the remaining DTMF signal as one or more DTMF packets. The egress VoIP gateway responds by transmitting the false audio and then transmitting the DTMF signal into the non-VoIP network. As a result, instead of sending a properly formed DTMF signal into an interfacing non-VoIP network, a short burst of energy that represents the false audio is sent followed by a well formed DTMF signal. DTMF detectors in the non-VoIP interfacing network, for example a detector used with a conference service to screen participants using DTMF-based passcodes, can fail due to the false audio that precedes the DTMF signal. Such DTMF detection failures can have significant and negative effects on customer satisfaction, and thus can be a major problem for VoIP service providers. For example corrupted DTMF signals can cause the teleconferencing service to reject the calling party on the basis of a non-valid passcode entry.